Let's formalize it a bit, I think this helps clarify.
Let A be the hiring propensity, B be job preparation (education, experience, what have you), and C be the "blinding factors" Helsinki is trying to address.
As a model we would like the joint propensity for everyone to be normalized. In other words, P(A,B,C) would be about the same.
The joint propensity P(A,B,C) would be expressed as
P(A,B,C) = P(A | B,C) * P(B | C) * P(C) [0]
Helsinki is making the first term, P(A | B,C) independent of their "blinding factors", or, in other words, P(A | B,C) = P(A | B).[1]
The remaining terms P(B | C) and P(C) are unaffected. It is P(B|C) differential that many arguing for what I call non-autarky "social corrections" (we do something as a society).
Since Helsinki's decision affects _only_ the conditional probability P(A| B,C), the other factors are still at play, and P(A,B,C) remains at issue.
There is nothing that can be done by Helsinki's hiring department for P(B|C) in the short term -- people are dealt the hand they have, life isn't fair, and so on. Completely true.
By blinding themselves, they set P(A|B) to be flat on C and no "equalization" of P(A,B,C) can be observed.
A truly egalitarian society would have P(B|C) = P(B). However, we aren't there. C, what Helsinki blinds themselves to, matters for qualifying life experiences.
The model that is being argued for by you and others I believe is formalized as a conditional independence model, namely that the propensity of hiring is independent of race, conditional on life preparation: P(A,C | B) = P(A|B) * P(C|B).[2] Note though: bias still persists: P(C|B) = P(B|C)*P(C)/P(B) -- we can't get away from that P(B|C) term in either model, and we know qualitatively that P(B|C) != P(B), hence why a correction can be positively argued for (rather than solely from emotion or duty).
Anyhow, that's the model in mind, and I'm open to being wrong/corrected. Thoughts?
--------------------
After thinking about it more, it seems that the ultimate goal of Helsinki is to make P(A|B,C) = P(A|B), not the latter model.
--------------------
[0] Chained conditional probabilites: https://en.wikipedia.org/wiki/Chain_rule_(probability)
[1] Independent variables definition: https://en.wikipedia.org/wiki/Independence_(probability_theo...
[2] Conditional independence: https://en.wikipedia.org/wiki/Conditional_independence